Nvidia has transitioned its Vera CPU systems into full production, marking a shift toward purpose-built processors for agentic AI. The first infrastructure units were hand-delivered by Nvidia executive Ian Buck to early enterprise adopters, showcasing the company’s commitment to integrating custom CPU solutions into its AI factory architecture.

Anthropic and OpenAI received initial Vera CPUs for scaling models that act and simulate autonomously, while SpaceXAI is evaluating the systems on reinforcement learning workloads and agent-based simulation pipelines. These deployments highlight the shift from simple AI responses to complex, action-oriented applications requiring high-throughput processing.

Oracle Cloud Infrastructure is set to become the first hyperscale cloud provider to deploy Vera CPUs, planning to roll out hundreds of thousands of units starting in 2026. This large-scale adoption underscores Nvidia’s strategy to embed its processor technology deeper into cloud infrastructure and capture a growing share of enterprise AI workloads.

Vera CPUs are a key element in Nvidia’s co-design ecosystem, working alongside Rubin GPUs, BlueField-4 DPUs, Spectrum-X networking and MGX rack architectures. As the host processor for the Vera Rubin NVL72 platform, Vera connects via second-generation NVLink-C2C to Rubin GPUs, enabling seamless data flow for next-generation AI training and inference.